Another Fluke 5440B/AF Repair

[essele]

Ok ... all tests now pass, both analog and high voltage. I still need to deal with the missing buttons on the front and a bit of tidy up (and a few more caps to do), but I think it's almost there as far as basic functionality goes!

Is the output voltage stable?

I've only just started to look at the output and there are lots of caveats to this ... the room is a bit chilly and I can feel cold draughts (although can't find a source) and I'm stringing two connected sets of cheap Chinese test leads across the room, neither the meter (DMM7510) or the 5440B are properly warmed up (probably about 90 mins at this point) ... ignore the absolute numbers, the 5440B isn't calibrated and the DMM7510 isn't self-cal'd ... so there is a huge set of error sources in this, however....

From the attached graphs you can see the output, this is from about 1 second after hitting OPER ... each time I run it the ultimate output seems very consistent and settles at 10.000417 (averaging over different time periods, using 5NPLC and no filter) although has been slightly rising in more recent attempts to 10.000419, I suspect this may be to do with things still warming up.

(Later note -- it's now at 10.000425 and has been consistently increasing, does look like a warm-up process)

The first graph (102055) is from a run where it wasn't at standby for very long, probably no more than 3 minutes ... it looks absolutely fine to me, the specs say it can take 10.5 seconds to settle to within 0.5ppm from STBY to OPER, and it's probably within 0.2ppm at that point.

The second graph (103148) is after a much longer STBY period, probably 10 minutes or so, and you can see a lower dip (from stable state), although still only about 0.6ppm, but it does then take a bit longer to stabilise which may well be to do with relay heating/equalising.

Either way I'm not sure I'm seeing a particular problem -- although I will look again once I've found less error-prone ways to measure this. To be honest I'm absolutely stunned how good this seems to be ... this thing has clearly been in a room with a fire, filled with smoke, filled with water, left to soak for a while, had all the really expensive relays replaced with different configuration low cost Chinese ones, and been messed with by a fairly incompetent idiot that only half knows what he's doing ... it's a miracle it's working at all!

[baof]

At the beginning of output, the curve is straight. In less than 1 minute, the coil heating begins to affect. The curve continues to decline until the coil heating temperature reaches a stable level

[essele]

I'm certainly not an expert but 30 minutes seems like quite a long time for small relays to heat up, what makes you think it's the relay?

Which model relays did you use?  I used relatively low power ones (< 200mW) ... if you used the standard power Panasonic ones (400mW) then I suppose that could account for some differences, with three relays in place of one original one, that's 1.2W, which is significantly higher than the original relays which I *think* are 450mW.

Have you changed all of the relays? I'm wondering if having a mix might introduce inconsistencies between output and sense perhaps? I'm not seeing this at all, but I've changed everything.

Have you also tried running the same experiment with different cooling-off times (i.e. time in STBY before switching to OPER) that may help to confirm it's heat related.

If you're concerned about a specific relay, and it's a 6-Pole one, then I can probably knock you up one of my PCB's with the LCSC relays on it (for cost + postage) if that's useful?

[Kleinstein]

There are several parts that can show a slow response. Thermal effects at the relays can be one - some 10 minutes may be about the right time scale, at least if there is no strong fan. There may be other parts that are effected from heat flows changing too, like some OPs and maybe the DAC part, though ideally the effect should not be large.
Another point is dielectric absorbtion in the capacitors at the filter of the PWM DAC.  This can also take quite some time to stabilize.
One would see the difference by comparing the turn on with a large step (e.g. 0.1 V to 10 V). The relays would be the same for 0.1 and 10 V. For the DA the start from 0.1 V should not make much difference from the off state.

[baof]

Thank you for your reply
I use Panasonic ds2y-s with a working current of less than 40mA. I replace all relays on A7. I tried to turn off the output for a short time, but the coil did not cool down. When I turn it on again, the output will not be affected. If the output is turned off for a long time, the coil will cool down, and the output will be affected. It is confirmed that it is the effect of relay heating. Maybe it is a perfect replacement. I really need to use magnetic holding relay.
I'm in China. I have two free opportunities to do PCB at JLC every month (even postage is not needed)

[Kleinstein]

Other effect like the fitler cap DA can have an effect on comparable time scale as the relay heating. So I don't think the simple test with a short and long turn off does proof that the drift is due to relay heating - it is well possible, but not for sure. At least the filter caps could be excluded, by comparing to steps in voltage, turning down the voltage to near zero only with the DAC.

It looks like they sometimes use multiple contacts in parallel. So there may no actually be a need to have so many contacts and in some places one could replace a 6 pole relay with just one 2 pole relay , e.g. leave some small relays on the adapter board unpolulated.

Magnertic latching relays are a bit tricky:  they need additional control circuit and more important the state after turn on is not so well defined. A similar problem can happen on power down if only some relays are replaced.  Wrong set relays may do damage to the circuit or give out dangerrous voltages.
Latching relays are more a thing to consider in a design from the start, nothing to change easy later.

[essele]

First attempt at a keycap from the front panel ... not bad for a first ever resin 3d print!

I still have a little tweaking to do on the shape ... I was convinced the front was a "squircle" but perhaps I'm over complicating things and it's just a rounded corner square. General dimensions are pretty close.

And obviously the legend isn't quite working yet, I have a few different ideas to try ... and I need to see if I can find a better font.

The next one will be in a CMYK pigment mixed resin, so I can start to see how close to the colours I can get.

[MegaVolt]

Here's an interesting video on how to make buttons. Maybe there will be something interesting there. The quality is great

[essele]

Wow .. they are stunning, lots of work and lots of jigs/tools needed. A very interesting video, but I'm not sure there's anything specific that would help my approach. I love the use of lego though!

Mine are getting better ... the shape is much closer now, but still needs a couple of tweaks, and the deeper legend cutout means more depth of colour on the legend which seems like it will work out, I get some discolouration caused by seepage into the tiny ridges on the print surface, I'm hoping they will polish out.

The pigments are a challenge ... one drop is not granular enough, but I'm reasonably close, I'm hoping adding a bit more white base resin will lighten this up sufficiently. (Although the resin doesn't seem to be 100% opaque so I don't think it will ever completely match.)

I found the right font for the smaller text (I'm using Blogger Sans Medium which seems to match pretty well), however the bigger font (i.e. the digits) seems to be a wider variety ... "Kingthings Clarity" looks reasonable, although I haven't tried it yet.

[essele]

Ok ... getting better, but still a few problems.

For the larger symbols I went with the same font but just expanded it horizontally, seems to look ok.

The colour doesn't quite match yet, it's a bit challenging because it darkens slightly during UV curing, but I've got some scales coming today which will hopefully help with more accurate mixing, and solve the one-drop is not granular enough problem.

The biggest issue is the lighter colours are not particularly opaque and when combined with a need to have a deep trough for the legend it means that the black is visible through the lighter colour making it look like it's smudged. I don't think this will be an issue on the darker ones. The solution may be a much shallower trough and using a different solution for the legend (i.e. not black UV resin.)

I've even managed to get a bit of a gloss finish going!

This is a bit silly really, but I'm enjoying it (apart from the resin fumes, which are evil!)

[MegaVolt]

It looks really good!

[ch_scr]

Agreed on it looking really good! How long do you take to do one cap? Because that's the easy way out on 'identical looking keycaps' - "just" manufacture them all...

[essele]

Thanks ch_scr.

They take just under an hour to print, although I could do several (in the same colour) at one time ... at the moment I'm still trying to get the colours right so it doesn't make sense to do more than one.

The "post processing" probably takes an hour as well ... clean, UV cure, sand a bit, do the legend, cure again, then lots of sanding/polishing.

My challenge is that pretty much all the keys I need are different ... only one light one (the number 7), a few brown ones all with different legends (well, 2 of them are blank) and then 1 greyish one. So there isn't really a mass-production option.

I'm really struggling with the opacity challenge ... I don't think they are going to be quite right, but hopefully they will be good enough. I've still got a few more things to try, and it's all good learning experience (apart from the fumes ... definitely need a better mask!)

[essele]

I think I've got this figured out now... the solution is acrylic paint markers!

For the legend, on the lighter colour I've gone for a slightly shallower recess, but then by using a light colour acrylic paint marker to start with (I had a very light pink which was the closest to the light resin) it puts a fully opaque coating in the recess, but it dries thin so doesn't fill the recess. Then you can cover with black and it doesn't have the smudgy appearance that was previously caused by the black being visible through the lighter colour.

Also, by polishing a bit more the appearance is more opaque ... so I'm going to go with the number 7 as shown ... just a little more polishing on the edges.

The acrylic paint markers also work brilliantly on the small text as you can see on the "limit" cap ... I have a few more experiments to hopefully improve the colour, and then I still need to do the grey/beige one.

See the photo of the front panel with the "7" ... it's not fully pressed in yet as it's pretty tight and I don't want to keep taking it off, but otherwise I don't think it looks too bad.

[MegaVolt]

If you hadn't told me I wouldn't have been able to find a different one

[essele]

I think I'm done ...

I've given up trying to match the brown key caps and created a nice dark grey which I think looks great. I'm not entirely happy with the "div" key, but I'm not doing it again, and one of the three menu buttons is slightly sticky so I should have replaced the switch for that ... I'll do it next time I open it up.

Other than that I'm pretty pleased with the front panel now -- before and after photos attached.

Everything is working nicely, on to some testing!

[rigrunner]

That's an excellent job. My hat is off to you. 

[MadTux]

Just milled a new display window for another cheap 5440B from ebay on my FP4NC.
Used dark grey 3mm plexiglass window from ebay ($20), looks better than new, IMO.

Unfortunately also missing the EMI wire screen, no idea if I should replace it with stainless/copper wire mesh from ebay or just glue it in place without EMI wire mesh??
Got a quote from some China company for new real EMI wire mesh, $280. Total ripoff!

Apart from that, that's my 4th fully functional Fluke 5440B 

[essele]

Very nice!

What was the reason for milling? Just a thickness difference? I don’t recall there being any indents on there.

[MadTux]

Reason for milling/routing: much quicker, easier and much more accurate than by hand with hacksaw and files ;-)

Somehow you can only get 3mm Acrylic, think the original window is 2mm, but not sure.
First plan was to mill the window down to 2mm, where it contracts the front frame, so the window would sit at a millimeter deeper than with 3mm Acrylic.
But when I tried it, the 3mm version was flush with the rest of the front panel, probably because I left out the EMI screen, so I just glued it in place.
No EMI screen shouldn't matter that much, Keithley 2000 and friends don't have them either...

[leighcorrigall]

While I was searching for other things, this came up and might be useful to people.


Datanetics DC-60 series:

https://deskthority.net/wiki/Datanetics_DC-60_series

[leighcorrigall]

...

To get at the fan (and mains filter) I've removed the back panel and the A19 Outside Guard Terminator PCA (which seems to be just a set of unregulated power supplies) ... on the board there was an area which was quite darkened, mostly around CR2, CR3, and F2. Some tracks on both sides were missing their solder mask and I think the plastic coating that would have been on CR2/CR3 appears also to have burnt off! F3 had some damage to the sheath underneath where it was close to a track which appears to have overheated.

All of the fuses are intact and don't look like they have been replaced (they do now because I messed with them to apply some new solder mask underneath -- and looking at the photo I need to redo my soldering!)

So I am a little confused ... CR2 and CR3 feed the unregulated +5V supply, both are fused with 5A fuses, both diodes appear to be ok. However something has caused them and the PCB tracks to get very hot but it hasn't blown the fuses! R2 measures fine, and C2 seems to be ok from an external timed charging test.

I'll obviously check the voltages once I get it all back into the chassis.

Any thoughts?  Is this just to be expected with age? Should I replace CR2/3 and the fuses?

This is probably the simplest board in there --- this is going to be a long journey ;-)

Thanks,

Lee.

Hi Lee,

I have the exact same issue with my board. The diodes seem to be the problem given the localization of the burn marks. On further inspection, no other boards on my unit appear damaged. The instrument passed its diagnostics and it can output 0 to 100 V without any noticeable problems so far. I have yet to test above 100 V because I have trust issues with antique equipment that has not been fully inspected. The unit was calibrated in 2018, so I imagine that it has passed some quality checks and this problem is nothing new.  My VAC was set to 115 at the factory and may cause some additional warming because my line power is 120 VAC. I'd say this instrument has been around the block.

Can you please tell me how you removed the outside guard terminator PCB? It looks complicated and I must be missing something so far.

Thank you.

EDIT: Solution found. See attached pictures of the removed outside guard terminator PCB.

[essele]

Hi Leigh,

It's been a while and I can't easily open it open again to refresh my memory, but it absolutely was a pain.

I think it's the screws on the back panel (can't remember if this is two or four) that hold the transformer in place, and I think you need to remove the rectangular metallic capacitor as well as it gets in the way. The transformer stays attached to the board and it all comes out ... there's an edge connector that it plugs into, so once you've unscrewed everything you need to it should just pull out (remembering to remove the connections to the other large cap.)

Mine looked very similar to yours ... just worse!

Lee.

[leighcorrigall]

Hi Leigh,

It's been a while and I can't easily open it open again to refresh my memory, but it absolutely was a pain.

I think it's the screws on the back panel (can't remember if this is two or four) that hold the transformer in place, and I think you need to remove the rectangular metallic capacitor as well as it gets in the way. The transformer stays attached to the board and it all comes out ... there's an edge connector that it plugs into, so once you've unscrewed everything you need to it should just pull out (remembering to remove the connections to the other large cap.)

Mine looked very similar to yours ... just worse!

Lee.

I figured it out based on what you said. The process is simple:

1) Remove the top cover of the instrument that is held in place with 8 flat head screws.
2) Discharge the two large capacitors that are near the outside guard terminator. I used a voltmeter set to the highest range. Disconnect them and cut the cable tie. Remove the capacitor connected to the rear panel with two screws.
3) Remove the two top screws on the rear axial fan after removing the fan filter. The bottom of the fan is held in place with a bracket. You will need to use a bit of force to remove the fan because it is tangentially spring-loaded on the screws.
4) Unfasten the outside guard terminator transformer mounted to the rear panel with four screws.
5) While making sure that the board doesn't short on the chassis, lift the outside guard terminator by the transformer. If you raise the card, the weight of the transformer will cause the card to unintentionally contact the metal chassis after it is removed from the card socket.

I do not have any image editing software installed on my computer. Sorry about only providing 3 images. These should be sufficient as references.

EDIT: See images of PCB in the previous reply.

[leighcorrigall]

There seems to be more thermal damage at the outside guard regulator (A17) on my unit.

Damage is related to R1 (0.15Ω, 2W, WWR?), Q1 (40251) and U5 (LM317HVK). R1 appears to have been replaced at some point but other components may have as well.

Edit: Someone else has the same issue with R1. URL: https://dabbledoo.weebly.com/fluke-5440b.html